Home

Awesome

<!-- PROJECT LOGO --> <p align="center"> <h1 align="center"> ESIM及VID2E (个人使用版本) </h1> <h3 align="center"> <a href="https://kwanwaipang.github.io/File/Blogs/Poster/esim.html">Blog</a> | <a href="https://github.com/uzh-rpg/rpg_vid2e">Original Github Page</a> </h3> <div align="center"></div> <br>

配置记录

<!-- 下载到当前目录 -->
wget https://rpg.ifi.uzh.ch/data/VID2E/pretrained_models.zip

<!-- 解压 -->
unzip pretrained_models.zip
<!-- 删除已有环境 -->
conda env list
conda remove --name vid2e --all

<!-- 遇到conda创建很慢 -->
conda config --show #看看channels
conda config --show channels
conda config --remove channels conda-forge

conda create --name vid2e python=3.9
conda activate vid2e
pip install -r requirements.txt
conda install pybind11 #此处不要安装matplotlib
<!-- 注意要指定一下pytorch的版本 -->
conda install pytorch==1.12.0 torchvision==0.13.0 torchaudio==0.12.0 cudatoolkit=11.3 -c pytorch

pip install esim_py/

pip install setuptools==69.5.1
pip install esim_torch/

测试是否安装成功以及debug

conda activate vid2e

<!-- 测试esim_py是否安装成功 -->
python esim_py/tests/test.py
#如果报错GLIBCXX_3.4.30
ln -sf /usr/lib/x86_64-linux-gnu/libstdc++.so.6 /home/gwp/miniconda3/envs/vid2e/lib/libstdc++.so.6
export LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libtiff.so.5
python esim_torch/test/test.py
采用ipynb,选择kernel为vide的
在服务器终端安装一下pip install ipykernel 会更快
pip install -r requirements.txt --force-reinstall
#然后重启ipynb的kernel,在终端安装更快:pip install ipykernel 
device=cpu
# device=cuda:0
python upsampling/upsample.py --input_dir=example/original --output_dir=example/upsampled --device=$device

测试记录

用HKU-dataset生成数据集同时与real event data 进行对比

pip install --extra-index-url https://rospypi.github.io/simple/ rospy rosbag

pip install rosbag_pandas

pip install cv_bridge

pip install sensor_msgs --extra-index-url https://rospypi.github.io/simple/

pip install geometry_msgs --extra-index-url https://rospypi.github.io/simple

用TartanAir生成event 数据集

参考资料